But, as can be confirmed by checking the score at IMSLP (conveniently linked from that page)… it has 5 parts, not 7. The first two works would appear to be the proper first part, split in two—so I suppose they could be merged. Parts 4 and 5 match up (though are named wrong/weird translations). “Supplications de Poiseau de feu” is just missing. #6 and #7 would appear to be from the 1919 suite (at least, from the names)
Looking at the linked recordings, it’s a mess—some of the recording titles imply they’re the ballet version, not the suite. Or say that they’re the 1919 version.
I’m going to re-enter the 1911 suite, clearly labeled as such, with an annotation noting there is more than one version. Note sure what to do with the existing mess, though.
According to Wikipedia some recordings of 1911 suite list 1st movement as three movements. Editors commonly create work parts based on track titles on releases but like now sheet music and majority of recordings split it differently. Guideline about this is clear: “Works should be entered into MusicBrainz using the boundaries decided by the composer, as printed in the score”. At least this time score is easily available at IMSLP.
Relationship between 1911 & 1919 seems to be incorrect because these aren’t different versions of the same work. These are different works which are both based on the ballet.
When editors add different versions of same work they should remember to add also catch-all work for unknown versions (when these are hard to identify). It seems many recordings of 1945 suite have been incorrectly linked with 1911 & 1919 suites.
One way of fixing this would be linking all parts which can’t be identified with “catch-all for unknown versions” work.
Immediately after your message I didn’t understand how bad the situation currently is. http://musicbrainz.org/work/d2a1f57e-f7c9-4098-a716-f5c72815e0ab is basically mix of multiple of works. I think it’s better not to merge it and it’s parts with anything but update disambiguation and make it “catch-all for unknown versions”.
First of all, the English, French, and Russian translations are missing as aliases (and maybe in disambiguation too). Then, maybe one language should be selected for every major version of this work for consistency, most probably from the score, just as for boundaries. Finally, regarding the existing mess, I second @ListMyCDs.com for turning it into a ‘catch-all for unknown versions‘ work entity along with an annotation that explains the mess and point to how to distinguish between them, as I did for Paddy Fahey’s for example (improvements are welcome).